2.2 Assimilation, accommodation, and cognitive development

Jean Piaget’s developmental epistemology provides a striking psychological precursor to this distinction. Piaget described adaptation through two coupled processes. In assimilation, experience is interpreted through an existing cognitive schema. In accommodation, the schema itself is modified because the existing organization cannot adequately incorporate the experience. Cognitive development proceeds through repeated attempts to restore an always provisional equilibrium between them [ Piaget , 1952 , 1985 ] .

In the notation above, assimilation is analogous to changing a realization while retaining its package,

whereas proto-accommodation may propose a package transition

Independent admission turns that proposal into accommodation. Only when the presentation or generative language changes does the admitted transition also contain a theory map \(J:\mathbb S\to \mathbb S^+\) and qualify as a theory extension. The analogy is not an identification of a child’s cognition with a categorical machine. Its value is architectural: learning cannot always be understood as adjusting content inside a fixed representational scheme. Sometimes the scheme that determines what counts as an object, an invariant, or an explanation must itself change.

Chapter 1 used conservation to show how an invariant can be acquired across changing appearances. Object permanence gives the complementary example: a theoretical object organizes appearances even while occluded [ Piaget , 1954 , Piaget and Inhelder , 1974 ] . Here we retain only their common architectural question: when should a reasoner vary a model of existing objects, and when does persistent failure warrant a new object, invariant, or law?

. Assimilation repairs a model within a maintained conceptual declaration. Proto-accommodation probes how that organization might have to change. Accommodation installs an admitted package change; theory extension is the case that changes its presentation or generative language. A creative system needs all three distinctions, together with evidence separating a genuine need for theory change from an ordinary failure of fit.

This is a conceptual lineage, not a categorical model of child development or a commitment to rigid stages. The computational problem is to make the boundary between the two operations explicit and auditable.

A computational precedent.

Gary Drescher made this connection to artificial intelligence explicit at the MIT Artificial Intelligence Laboratory. His early thesis introduced the Schema Mechanism; the subsequent paper, also issued as an AI Memo, was aptly titled “Genetic AI—Translating Piaget into Lisp” [ Drescher , 1985 , 1986 ] . The mechanism begins with sensorimotor primitives and learns predictive schemas organized as context, action, and result. It can refine schemas in response to experience, compose actions, and construct new representational items when its existing vocabulary is inadequate.

Drescher’s 1989 doctoral thesis was later published as Made-Up Minds: A Constructivist Approach to Artificial Intelligence [ Drescher , 1989 , 1991 ] . Its simulated infant supplied an executable test of whether a schema-building mechanism could reproduce parts of Piagetian sensorimotor development, including steps toward a persistent object concept. In the language of this book, the Schema Mechanism is an early computational architecture in which assimilation updates predictive structure inside an available vocabulary, while accommodation can appear as the construction of a new item or coordinated schema.

The resemblance to the synthetic-creativity program is substantive but incomplete. Drescher supplied an embodied microworld, an explicit action loop, and a mechanism for concept invention—all features absent from purely textual generation. The present program asks how this constructivist approach can be generalized from sensorimotor schemas to typed algebraic theories, active scientific experiments, explicit theory transport, and independent admission.